I have a paper coming out in Econometrica this month (ungated version here) on how local Islamist mayors increased female participation in education in Turkey during the 1990s. The main contribution of the article is that it uses a method called the Regression Discontinuity Design (RDD) which allows identifying an estimate of the causal effect of having Islamist political control.

A focus on causality is important as the reason why Islamist-controlled constituencies have poor women’s rights may come from voter-specific characteristics, like preferences for political Islam, as much as the effect of the Islamist politician itself. If politicians have little power to influence policy other than adopting the policy position of a representative (often the median) voter, then (Islamic) party identity doesn’t matter for policy, only voter preferences do. Resulting policies may be detrimental to development, but is then the result of voter preferences, not politicians. If for some reason, elected politicians can influence policy away from the representative voter to their own preferred position, then politician identity will matter for policy. In this case, unwanted policies may be as much the fault of the politician (see this paper for more through discussion). Continue reading →